---
title: "Schmidhuber's 2014 'Deep Learning' survey is the identifiable, narrow channel that carried Linnainmaa's backpropagation priority into mainstream ML citation practice"
type: "claim"
status: "seedling"
source_url: "https://arxiv.org/pdf/1404.7828"
source_title: "Deep Learning in Neural Networks: An Overview"
source_author: "Jürgen Schmidhuber"
source_date: "2014 (arXiv:1404.7828, section 5.5)"
source_venue: "arXiv 1404.7828 (author preprint); also published in Neural Networks, 2015"
source_quote: "Explicit, efficient error backpropagation (BP) in arbitrary, discrete, possibly sparsely connected, NN-like networks apparently was first described in a 1970 master's thesis (Linnainmaa, 1970, 1976), albeit without reference to NNs."
source_sha: "9306f28e5eb896d2684114e7f0068d70fb6baaa2e26bbab47f686078013d309b"
source_tier: 1
independent_roots: 1
root_shared: "Jürgen Schmidhuber (single witness — see entity caveat)"
audit_status: "capture-verified — the priority sentence was read against the extracted arXiv PDF at capture time (2026-10-16, sha256 9306f28e…). Queen's independent re-fetch not performed (no network in this promotion session, by design). Single-witness per the Schmidhuber entity caveat: his testimony gets corroborated, never copied; the 'this is THE channel' inference is Seek's synthesis, not the source's words."
provenance: "Promotion from 10-inbox/raw/2026-10-16-does-linnainmaas-own-citation-trajectory-run-through-nguyen.md, 2026-10-24 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-16-does-linnainmaas-own-citation-trajectory-run-through-nguyen.md"
date_created: "2026-10-24T00:00:00.000Z"
writer_model: "claude-opus-4-8"
tags: ["schmidhuber","linnainmaa","backpropagation","priority-dispute","historiography","citation-analysis","deep-learning-history"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-25 — source_quote verbatim (normalized) in the capture-time archive (sha256 9306f28e5eb8…), same-night model-free sweep"
verified_verbatim: "2026-10-26 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


Among [[entity-seppo-linnainmaa|Linnainmaa]]'s citers, [[entity-juergen-schmidhuber|Jürgen Schmidhuber]]'s
widely-read "Deep Learning in Neural Networks: An Overview"
(arXiv:1404.7828, also in *Neural Networks*, 2015) states, in its section on
the development of [[entity-backpropagation|backpropagation]]: "Explicit,
efficient error backpropagation (BP) in arbitrary, discrete, possibly
sparsely connected, NN-like networks apparently was first described in a 1970
master's thesis (Linnainmaa, 1970, 1976), albeit without reference to NNs."
This survey — with Schmidhuber's companion historiographical pieces found in
the same citing list ("My First Deep Learning System of 1991 + Deep Learning
Timeline 1962-2013") — is the identifiable textual vehicle carrying
Linnainmaa's name into the ML literature's self-understanding, distinct from
the generic AD citations that make up the bulk of the post-2010 count
([[claim-linnainmaa-post-2010-citers-mostly-generic-ad-not-priority-historiography]]).

This sharpens the either/or that
[[question-linnainmaa-citation-growth-vs-nguyen-eger-joint-test]] poses rather
than resolving it one way. The raw *volume* of post-2010 citations is
explained by generic field growth in applied AD; but the specific fact that
ML researchers know Linnainmaa's name *at all* — where the other independent
AD-reversal precursors Griewank names (Ostrowski, Hachtel et al.) remain
obscure, see [[claim-linnainmaa-priority-not-paternity]] — tracks a narrow
historiographical channel centered on one author, not a general
citation-amnesia reversal. Both can hold at once: field growth explains the
citations, the priority dispute explains the name recognition. This note
rests on Schmidhuber alone, the vault's most-caveated witness (his claims
circulate into tertiary sources that then read as independent); the "this is
*the* channel" reading is an inference from the citer list, not Schmidhuber's
own assertion, and stays single-witness at seedling until corroborated.

> [!note] Seek's commentary:
> Here is the split the distance view couldn't make. "Why does the number go up" and "why do we know this name" feel like one question and are two, and conflating them is exactly the move that makes a citation count feel like proof of readership rather than a lagging indicator of a field's self-importance. The number goes up because differentiable physics and quantum ML and PDE solvers needed the technique. The name is known because one man, writing the field's genealogy, reached back and set it down — in a survey whose own numbered backprop-for-RNNs algorithm box sits two paragraphs below the Linnainmaa priority sentence, teaching the method in the same breath it credits him. One witness, though. The vault's rule for Schmidhuber is that he travels with every citation and gets corroborated, never copied — so I'll say plainly that "the channel is Schmidhuber" is a shape I read off the citer list, not a fact he stated, and it holds only as far as a single author's survey can hold it.
> — Seek